Re: severely bad timekeeping

[Donald Lee <MacPPC2%caution.icompute.com@localhost>]

I found some notes in my logbook for my older machines.

The clock "base frequency" is based on the "timebase-frequency" in OF.

I saw this same kind of time drift with the 7600+G3-upgrade combination,
and fixed it by adding the following to open firmware:

        setenv boot " /PowerPC,604" select-dev A7D8c0 encode-int " 
timebase-frequency" property boot

The "timebase-frequency", as I recall, is not something that is in the NVRAM,
so you can't just change it.

I had to do several newtonian iterations to get this close to correct before
NTP would work.  When the drift is too large, NTP gets very unhappy,
and just gives up.

The calculation that I used was to simply take the percent time drift and
multiply the "0xA7d8C0" constant by the error.  This got me pretty close.
The default, as I recall, was something "round", like 0xA80000.

There is more info in and around the following list thread:

        http://mail-index.netbsd.org/port-macppc/2000/03/14/0012.html

Enjoy,

-dgl-

At 12:53 AM -0400 4/17/09, der Mouse wrote:
>I've got a macppc machine running 4.0.1.  ntp doesn't work right, and
>I'm wondering whether this is a problem with my particular machine or
>with this kind of machine (FWVO "kind"), and whether there's anything I
>can do to fix it.
>
>The machine started life as a 7600/132 - for example, that's what's
>printed on the case - but (I'm told) it's got a G3 upgrade in it.
>dmesg.boot describes the CPU thus (I am not comptent to say whether
>this means it really is a G3 or not, which is why I put it that way):
>
>cpu0 at mainbus0: 750 (Revision 2.2), ID 0 (primary)
>cpu0: HID0 8090c0a4<EMCP,DOZE,DPM,ICE,DCE,SGE,BTIC,BHT>, powersave: 1
>cpu0: 305.89 MHz L2 cache present but not enabled 
>
>The time it keeps is wrong by almost a percent - it runs approximately
>half a second per minute slow.  ntp fails to sync under these
>circumstances (yes, I did check for an old ntp.drift file; this is
>being tested after destroying ntp.drift).  Killing ntpd and running
>ntpdate -b once a minute produces matching results (a correction of
>about .48 seconds each time around).
>
>Any ideas?  I can provide more details, if you can tell me where to
>find them....
>
>In passing, what would I need to do to figure out what L2CR_CONFIG
>settings I want for this CPU?  "L2 cache present but not enabled"
>sounds like a performance killer, and the macppc FAQ points to list
>mail indicating that setting L2CR_CONFIG is the right magic, but
>doesn't give much guidance on what to set it to.
>
>/~\ The ASCII                            Mouse
>\ / Ribbon Campaign
> X  Against HTML               mouse%rodents-montreal.org@localhost
>/ \ Email!          7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B